Human N-Myristoyltransferase Amino-terminal Domain Involved in Targeting the Enzyme to the Ribosomal Subcellular Fraction*

N-Myristoyltransferase (NMT) catalyzes the cotranslational acylation with myristic acid of the NH2-terminal glycines of a number of cellular and viral proteins. Most of the in vitro NMT activity (60–85%) in isoosmotic cell homogenates of human lymphoblastic leukemia (i.e. CEM and MOLT-4) and cervical carcinoma (i.e. HeLa) cells was shown to be associated with the ribosomal subcellular fractions by differential centrifugation. Also found in the ribosomal fractions was a ≈60-kDa protein that was specifically immunoblotted with an anti-human NMT (hNMT) peptide antibody. This ≈60-kDa protein was stable in the presence of proteolytic enzyme inhibitors but was gradually converted into a ≈46-kDa species when stored in the absence of protease inhibitors. Sucrose density gradient centrifugation of the ribosomal fraction resulted in the hNMT activity sedimenting exactly coincident with the 260 nm absorption profile and exhibitingA 260/A 280 absorption ratios >1.8, indicating an association of NMT with putative ribosomal particle(s)/subunit(s). The subcellular targeting of hNMT was also examined by immunoblotting subcellular fractions from HeLa cells transfected with plasmids containing FLAG epitope-tagged hNMT inserts corresponding either to the originally assigned hNMT gene or to an alternative open reading frame initiated from an in-frame start site upstream from the assumed hNMT start site. Anti-FLAG immunoblotting of cells transfected with a plasmid containing the larger insert revealed FLAG-NMT primarily in the ribosomal fraction with an apparent molecular mass similar to the ≈60-kDa native hNMT. In contrast, immunoblotting of cells transfected with a plasmid containing the smaller insert identified a ≈50-kDa FLAG-NMT predominantly in the cytosolic fraction. An analysis of mixtures of CEM ribosomes and serial dilutions of purified recombinant FLAG-NMTs demonstrated that the ≈60-kDa FLAG-NMT binds ribosomes with higher affinity than the ≈50-kDa FLAG-NMT. These in vivo and in vitrosubcellular targeting and recombinant expression experiments identify a native hNMT that is 10–12 kDa larger than the enzyme predicted by the originally assigned hNMT gene and which is apparently translated from an alternative up-stream start site. The data also indicate that although the unique NH2-terminal residues encoded by this larger open reading frame are not required for in vitrocatalytic activity, they do provide signal(s) involved in targeting hNMT to the ribosomal subcellular fraction where cotranslationalN-myristoylation occurs.

The amino groups of NH 2 -terminal glycines of a number of proteins that are essential to normal cell functioning and/or are potential therapeutic targets are found to be N-acylated with the 14-carbon fatty acid, myristate (e.g. ␣ subunits of heterotrimeric G proteins (1,2), GTP-binding arf1 (3), human immunodeficiency virus gag and nef proteins (4,5), the MARCKS (myristolated alanine-rich C kinase substrate) protein kinase C substrate (6), the protein phosphatase calcineurin B (7), the pp60 src protein tyrosine kinase (8), the retinal calcium-binding recoverin (9), the caveolae-associated endothelial nitric oxide synthase (10), the catalytic subunit of cAMP-dependent protein kinase (11), and mitochondria-associated cytochrome b 5 reductase (12)). N-Myristoylated proteins are therefore found associated with a variety of organelles with the myristate moiety required for such diverse functions as specific protein-protein or protein-lipid interactions, ligand-induced protein conformational changes, and/or correct subcellular targeting. This protein modification occurs almost exclusively cotranslationally within synthesis of the first 100 amino acids and is catalyzed by the enzyme myristoyl CoA:protein N-myristoyltransferase (NMT) 1 (EC 2.3.1.97) (13)(14)(15). Immunofluorescence microscopy reveals NMT to be distributed uniformly throughout the cytoplasm of yeast and mammalian cells (16,17). This finding plus evidence that N-myristoylation occurs on nascent polypeptides bound to free polyribosomes establish that NMT is physically localized and functionally active in the cell cytoplasm (14,18,19). However, no evidence has been offered identifying a more specific subcytosolic localization.
Protein N-myristoylation appears to be a tightly regulated reaction involving the coordinated participation of several different enzymes/proteins (e.g. N-methionylaminopeptidase, fatty acid synthetase, long chain acyl-CoA synthetase, acyl-CoA-binding proteins, etc.), access of NMT to pools of myristoyl-CoA, and the timely N-myristoylation of nascent polypeptide substrates to avoid potential interfering reactions (e.g. N-acetylation and polypeptide folding) (14, 20 -22). The ability of NMT to function in such a process implies the existence of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF020500.
‡ To whom correspondence should be addressed: mechanisms designed to ensure targeting of the enzyme to the appropriate protein synthesis machinery, possibly involving interactions with other cooperating components that facilitate the recognition and efficient N-myristoylation of the rapidly growing polypeptide substrates. However, no information is available regarding mechanisms regulating either the specific association of NMT with the cytoplasmic protein synthesis machinery (i.e. ribosomes, ribosome-associated factors, etc.) or even its direct participation during protein synthesis. Both are presumably required for NMT to accomplish the cotranslational N-myristoylation of proteins in mammalian cells.
Previous observations have raised questions as to whether NMT functions as a monomer or as part of a larger protein synthetic complex (23,24). It is also not clear if NMT functions in vivo as a soluble enzyme (or complex) or whether it acts in close association with insoluble subcellular structures (17, 22, 24 -28). For example, NMT activity is prepared as a soluble monomer from extracts of human erythroleukemia cells (29) and bovine spleen (30), but substantial amounts of the enzyme are found associated with the 100,000 ϫ g sedimented pellets in hypotonic extracts of yeast, rat (31), and bovine brain, 2 human, rat, and rabbit colon (27,32), and rat liver (27). The ready extraction of most of the NMT activity from insoluble pellets obtained from yeast, rat brain, and bovine brain with one or two sequential washes of low ionic strength buffers prompted the comparison of NMT with loosely bound peripheral membrane proteins (31), whereas its in vitro activation by various organic solvents prompted suggestions that it behaved as an integral membrane protein (25,27,33).
Also unresolved is whether NMT subunit sizes ranging from 48 to 67 kDa (26,29,30,34,35) are encoded by ORFs analogous to the originally assigned hNMT gene (36) (i.e. predicting a Ϸ48-kDa enzyme) and/or to ORFs comparable to the alternative in-frame start site(s) found Ϫ183 and Ϫ186 nucleotides upstream of the assigned hNMT gene (i.e. predicting a Ϸ55-kDa enzyme). Immunoblotting with anti-hNMT antibodies has identified a Ϸ60-kDa polypeptide as the predominant NMT subunit in bovine brain and human HeLa cells (17,24). Also identified were smaller immuno-cross-reacting forms (i.e. 46 -57 kDa) which apparently resulted from postextraction proteolysis of the NH 2 terminus of the Ϸ60-kDa bNMT subunit. A close relationship was also observed between NH 2 -terminal proteolysis and the conversion of high molecular mass forms of native bNMT (i.e. Ϸ120 and Ϸ400 kDa) into a fully active Ϸ50-kDa enzyme. This latter correlation suggested that the NH 2 -terminal domain of native bNMT, although dispensable for catalytic activity, may have in vivo regulatory functions related to subunit multimerization and/or NMT interaction(s) with other cellular proteins (24). A catalytically active 46-kDa NMT has also been purified from Drosophila which apparently results from NH 2 -terminal proteolysis of a larger precursor (37).
In the present study we show a specific association of hNMT with the ribosomal subcellular fractions from human lymphoblastic leukemia (i.e. CEM and MOLT-4) and human cervical carcinoma (i.e. HeLa) cells. We also use HeLa cells expressing FLAG epitope-tagged recombinant hNMTs (i.e. Ϸ60 and Ϸ50 kDa) corresponding to the two possible ORFs found in the hNMT gene (36) to demonstrate that the recombinant Ϸ60-kDa polypeptide translated from the larger ORF is similar in size to native hNMT and is localized to the ribosomal fraction, whereas the substantially smaller Ϸ50-kDa recombinant polypeptide is found primarily in the cytosolic fraction. Also shown is a preferential binding of recombinant Ϸ60-kDa FLAG-NMT to isolated CEM ribosomes. Our experiments thus provide in vivo and in vitro subcellular targeting and recombinant protein expression data indicating that the native Ϸ60-kDa hNMT is translated from a start site up-stream from the previously assigned hNMT gene and that the additional NH 2terminal residues included in the larger polypeptide are involved in the specific targeting of hNMT to the ribosomal subcellular fraction.

EXPERIMENTAL PROCEDURES
Cell Cultures-Cells were obtained from the American Type Culture Collection (Gaithersburg, MD) and included CEM (ATCC CCL 119) and MOLT-4 (ATCC CRL 1582) cells that were cultured in RPMI 1640 containing 10% fetal bovine serum and HeLa (ATCC CCL 2) cells that were cultured in minimal essential medium with 2 mM glutamine and 10% fetal bovine serum. All cell lines were grown at 37°C in an atmosphere containing 5% CO 2 .
Subcellular Fractionation-Mid-log CEM or MOLT-4 cells (1-3 ϫ 10 8 ) exhibiting a viability Ͼ95% by trypan blue dye exclusion were harvested and washed once with Tris-HCl-buffered saline, pH 7.4, by centrifugation at 480 ϫ g for 10 min at room temperature. All subsequent manipulations were carried out at 4°C. The washed cell pellets were resuspended in 4 ml of isoosmotic homogenization buffer A (i.e. 10 mM Tris-HCl, pH 7.5, containing 0.25 M sucrose and proteolytic enzyme inhibitors (i.e. 1 mM EDTA, 100 M TLCK, 4 M leupeptin, 0.3 M aprotinin, 20 g/ml soybean trypsin inhibitor, 110 M PMSF, 140 M TPCK, and 1 M pepstatin A)) or isoionic homogenization buffer B (i.e. 10 mM Tris-HCl, pH 7.5, containing 0.15 M KCl and the proteolytic enzyme inhibitors described above). Cells were disrupted with 25-35 passes of a tight pestle in a glass Dounce homogenizer until Ͼ90% of cells were broken as determined by trypan blue exclusion. One 530-cm 2 plate of HeLa cells (Ϸ90% confluent) was washed twice with Tris-HClbuffered saline, pH 7.4, and incubated on ice in 7 ml of homogenization buffer A (without sucrose) for 20 min. Cells were scraped from the plates directly into the Dounce homogenizer, and powdered sucrose was added to a final concentration of 0.25 M. The cells were then subjected to 25 passes in the tight Dounce homogenizer, resulting in Ͼ90% of the cells being disrupted as monitored by trypan blue dye exclusion. Subcellular fractionation of CEM, MOLT-4, and HeLa cells was accomplished by sequential differential centrifugation (38) using an SS-34 rotor in a Sorvall RC5B centrifuge (NEN Life Science Products) yielding pellets of nondisrupted cells and cell nuclei (480 ϫ g ave for 20 min), mitochondria (4,000 ϫ g ave for 20 min), and microsomes (20,000 ϫ g ave for 30 min) followed by centrifugation of the microsomal supernatant fraction at Ն 100,000 ϫ g ave for 90 min using a Ti 50 rotor in a L5-50 Beckman ultracentrifuge and yielding the ribosomal pellet and supernatant cytosolic fractions. Each pellet was washed once by resuspension in the original homogenization buffer, recentrifuged under the respective conditions, and finally resuspended in the same buffer. Assays for NMT and organelle marker enzymes were performed on freshly prepared subcellular fractions as described below. NMT was also visualized by immunoblotting proportional aliquots from each subcellular fraction with an affinity-purified anti-peptide hNMT pAb (24).
Isokinetic Sucrose Density Gradients-Constant velocity isokinetic gradients (5 ml) were prepared from 15% and 32% sucrose solutions in 10 mM Tris-HCl, pH 7.5, containing 25 mM KCl and the proteolytic inhibitors described above (39). The volumes of sucrose solutions and the centrifugation times were calculated for eukaryotic ribosomes (density Ϸ1.4 g/ml) (40). The physical setup for gradient formation was as described previously (41,42). The postmicrosomal supernatant fraction (1 ml) from a CEM cell homogenate was layered onto the preformed gradient and centrifuged at 50,000 rpm in an SW 50.1 rotor (Beckman) for 1.5 h at 4°C. After centrifugation, the tube was punctured at the bottom with an ISCO Tube Piercer (Lincoln, NE), and fractions were collected at 0.5 ml/min from the top of the gradient by displacement with a 55% sucrose solution. Fractions (0.25 ml) from the gradient were assayed for NMT activity, measured for absorbance at 260 nm and 280 nm, and sucrose concentrations were determined with a refractometer.
hNMT Expression Plasmids-cDNAs for hNMTs corresponding to either one or the other of the two possible ORFs in the hNMT gene (36) were obtained by pfu DNA polymerase (Stratagene, La Jolla, CA) polymerase chain reaction amplification of a reverse transcribed cDNA library derived from human cerebral brain mRNA (CLONTECH, Palo Alto, CA) using the following synthetic nucleotide primers: GCGCGCG-CAATTCATGATGGAAGGGAACGGGAAACG (i.e. Ϸ60-kDa hNMT) or GCGCGCGCAAGATCTTTATTGTAGCACCAGTCCAACCTT (i.e. Ϸ50-2 C. J. Glover and R. L. Felsted, unpublished observations. kDa hNMT) with 5Ј-EcoRI restriction sites as the sense primers and GCGCGCGCAGATCTTTATTGTAGCACCAGTCCAACCTT with a 5Ј-BglII restriction sequence as the common antisense primer. The resulting cDNAs were cloned into the pFLAG-2 bacterial expression plasmid (Kodak-IBI, New Haven, CT) to yield constructs expressing aminoterminal FLAG epitope-tagged hNMTs of Ϸ60 and Ϸ50 kDa, respectively (i.e. pFLAG-2-hNMTs). Inserts from the pFLAG-2 constructs were subsequently digested with HindIII and BglII and cloned into the pFLAG-CMV-2 eukaryotic expression vector (Kodak-IBI) yielding constructs expressing similar amino-terminal FLAG epitope-tagged hN-MTs (i.e. pFLAG-CMV-2-hNMTs). Insert sense and antisense strands were sequenced using dye primers or dye terminators and AmpliTaq FS with an Applied Biosystems Inc. Autosequencer. The nucleotide sequence of the inserts predicted the same amino acid sequence as reported previously except for a glutamine residue instead of a histidine at residue Ϫ4 from the originally published methionine start site (36). This glutamine for histidine substitution was in agreement with the GenBank entry (accession no. M86707) for myristoyl-CoA:protein Nmyristoyltransferase submitted by the original authors and was confirmed by sequence analysis of a cDNA clone obtained by 5Ј-RACE analysis described below.
Transfections-HeLa cells were seeded at 300,000 cells/100-mm tissue culture plate. After Ϸ24 h, 80% confluent monolayers were washed twice with Opti-MEM reduced serum medium (Life Technologies, Inc.) containing 200 mg/liter CaCl 2 . Cells were transfected essentially as suggested by the manufacturer with 16 g of the pFLAG-CMV-2-hNMT plasmid DNA and 70 l of LipofectAMINE reagent (Life Technologies, Inc.) in 12 ml of the Opti-MEM/100-mm plate. After 6 h at 37°C, the DNA and liposome-containing medium was removed, and 12 ml of fresh minimal essential medium containing 2 mM glutamine and 100 units/ml penicillin/streptomycin was added. After 24 h, the plates were rinsed twice in Tris-HCl-buffered saline, pH 7.4. 500 l of homogenization buffer A (without sucrose) was added, and plates were left on ice for 20 min. Cells were subsequently scraped from the plates, powdered sucrose was added to 0.25 M, the cells subjected to 25 passes through a "tight" Dounce homogenizer, and the homogenate processed by differential centrifugation as described above except that the postmicrosomal supernatant fraction was centrifuged for 30 min at 230,000 ϫ g using a 100.1 rotor in a Beckman TL 100 ultracentrifuge.
Expression and Affinity Purification of Recombinant hNMTs-Escherichia coli (strain BL-21/DE3) (Novagen, Madison, WI) transformed according to the manufacturer's recommendations with pFLAG-2-hNMT plasmids containing inserts corresponding to one or the other of the two hNMT gene ORFs were cultured in LB broth supplemented with 0.4% glucose and 100 g/ml ampicillin at 30°C with shaking until the A 600 reached about 0.4, treated with 500 M isopropyl-1-thio-␤-Dgalactopyranoside, and the shaking continued for 3 h. Recombinant proteins were extracted from bacteria by a procedure similar to that recommended by the plasmid manufacturer. Except where noted, all of the following operations were carried out at 4°C. Bacteria from 100 ml of medium were sedimented by centrifugation at 5,000 ϫ g for 10 min at 10°C and broken in lysis buffer (i.e. 50 mM Tris-HCl, pH 8.0, containing 5 mM EDTA, 0.02% sodium azide, 0.25 mg/ml lysozyme) at 12,000 p.s.i. using a French press at room temperature. Immediately after extraction the lysis mixture was supplemented with 0.1 volume of a 10 ϫ concentrated extraction solution (i.e. 2.0 M NaCl, 0.1 M CaCl 2 , 0.1 M MgCl 2 , 50 g/ml ovomucoid protease inhibitor, 167 g/ml soybean trypsin inhibitor, 300 g/ml TLCK, 17 g/ml each of leupeptin and aprotinin, 210 g/ml phenylmethylsulfonyl fluoride, 420 g/ml TPCK, and 6 g/ml pepstatin A) and centrifuged at 25,000 ϫ g for 1 h in an SS-34 rotor. Recombinant proteins were purified by applying the clarified supernatant solution to a Q-Sepharose anion exchange column (2.5 ϫ 16 cm) (Pharmacia Biotech Inc.) equilibrated in 50 mM Tris-HCl, pH 8.0, containing 0.2 M NaCl, 5 mM EDTA, 17 g/ml each of leupeptin and aprotinin, 167 g/ml soybean trypsin inhibitor, 300 g/ml TLCK, and 50 g/ml sodium azide followed by washing the column with the same buffer. The flow-through from the Q-Sepharose column was concentrated by ultrafiltration using a YM-10 membrane (Amicon, Beverly, MA), applied to a 5-ml anti-FLAG M2 affinity gel column (Kodak-IBI) equilibrated in Tris-HCl-buffered saline, pH 7.4, containing 1.7 g/ml leupeptin and aprotinin, and allowed to mix end-over-end for 25 min at room temperature. The settled gel was then washed with equilibration buffer at room temperature until A 280 reached background, and the recombinant FLAG-NMT was eluted with equilibration buffer containing 120 g/ml FLAG peptide. The FLAG-NMT was reequilibrated in 50 mM Tris-HCl, pH 8.0, containing 5 mM EDTA and 50 g/ml sodium azide by repetitive (3 ϫ) concentration to Ͻ1 ml and redilution with 10 ml of reequilibration buffer by ultrafiltration using a YM-10 membrane (Amicon) and finally concentrated with a Centricon 10 (Amicon). Affinity-purified FLAG-hNMTs exhibited single bands on SDS-PAGE by protein staining and by anti-hNMT or anti-FLAG immunoblotting and had similar specific activities (i.e. 1,500 -1,800 nmol/min/mg of protein) when assayed for NMT activity. 2 The final concentrated enzyme was stored in aliquots at Ϫ80°C.
Immunoblotting-Immunoblotting with an affinity-purified rabbit anti-peptide pAb to residues 27-38 (i.e. KTMEEASKRSYQ) of hNMT was described previously (24). Immunoblotting with anti-FLAG M2 mAb (Kodak-IBI) was similar except for incubation with the primary antibody for 30 min at room temperature. Immunoreactive bands were visualized using goat anti-rabbit or goat anti-mouse secondary antibodies conjugated to alkaline phosphatase, and 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium was used as substrate (Promega Corp., Madison, WI). Immunospecificity was assessed in duplicate lanes by preincubating the primary antibody with 20 M peptide antigen for 12 h at 4°C before blotting (24).
In Vitro Binding of FLAG-NMTs to CEM Ribosomes-The postmicrosomal supernatant fraction obtained by differential centrifugation of homogenates of 2.7 ϫ 10 8 CEM cells was separated into 580-l aliquots (i.e. equivalent to 45 ϫ 10 6 cells). To each aliquot was added 20 l of 2-fold serial dilutions of either the Ϸ60or Ϸ50-kDa affinity-purified FLAG-NMTs to yield the following final recombinant protein concentrations: 0.667, 0.333, 0.167, 0.0833, 0.0417, and 0.0208 g/ml. The mixtures were incubated at room temperature for 10 min and then centrifuged at 230,000 ϫ g for 30 min with a 100.1 rotor in a TL 100 ultracentrifuge at 4°C. Ribosomal pellets were finally suspended in SDS-sample buffer, heated for 5 min at 95°C, and analyzed by SDS-PAGE and immunoblotted with an anti-FLAG M2 mAb.
Assays-NMT activity was assayed essentially as described (31, 43) except for the presence of 0.1% bovine serum albumin. Assays for succinate dehydrogenase for mitochondria and NADPH cytochrome c reductase for microsomes were performed as described (44). Assays for lactate dehydrogenase for the cytosolic fraction and 5Ј-nucleotidase for plasma membranes were performed using diagnostic kits obtained from Sigma. The enrichment of ribonucleoprotein in the ribosomal fraction was assessed by A 260 /A 280 and A 235 /A 280 absorption ratios (45,46).
Analytical Procedures-SDS-PAGE (47) was performed on 10% polyacrylamide gels (Integrated Separation Systems, Inc., Hyde Park, MA or Novex, Inc., San Diego, CA), and proteins were visualized by Pro-Blue staining (Integrated Separation Systems). Prestained molecular weight markers were from Bio-Rad. Partially purified bNMT was prepared through the Sephacryl S-100 step as described previously (24). Protein concentrations were determined by the BCA method (Pierce) with standard curves constructed with bovine serum albumin. Synthetic peptides were purchased from Peptide Technologies (Washington, D. C.), and synthetic oligonucleotides were obtained from the Midland Certified Reagent Co. (Midland, TX). Peptide radioiodination was performed as described previously (31). 5Ј-RACE analysis (48) was performed using the Marathon cDNA amplification kit (CLONTECH) using a cDNA library derived from human brain mRNA (CLONTECH) and internal primers (i.e. GGGGTAGAACCAGTGCTCCACCTCCTC-CTGGC (nucleotides 860 -891) as the antisense and GAGCTGTTCT-CAGTGGGTCAGGG (nucleotides 49 -72) as the sense) to the hNMT gene (36). 3Ј-A overhangs were added to the full-length 5Ј-RACE cDNA by brief treatment with Taq DNA polymerase (PE Applied Biosystems), the cDNAs were TA cloned into pCRII (In Vitrogen, San Diego, CA), and the inserts were sequenced by Lofstrand Laboratories Limited (Gaithersburg, MD). Sigmoidal regression curve fitting to a three-parameter Chapman model was carried out using SigmaPlot version 4.0 (SSPS, Inc., Chicago).

RESULTS
Subcellular Localization of hNMT-Despite extensive evidence supporting a direct participation of NMT in the cotranslational N-myristoylation of nascent polypeptides (14,18,19), no experiments have been published demonstrating a specific association of NMT with ribosomes, nor has a mechanism(s) accounting for the enzyme's specific participation with the protein synthesis machinery of the cell been described. We have now examined the intracellular localization of NMT in human lymphoblastic leukemia (i.e. CEM and MOLT-4) and cervical carcinoma (i.e. HeLa) cells. CEM cell extracts were prepared in homogenization buffer A containing 0.25 M sucrose and separated into nuclear (480 ϫ g/10 min), mitochondrial (4,000 ϫ g/20 min), microsomal (20,000 ϫ g/30 min), ribosomal (Ն100,000 ϫ g/90 min), and cytosolic supernatant (Ն100,000 ϫ g) fractions by differential centrifugation. Each subcellular fraction was assayed immediately for subcellular organelle marker enzyme activities (i.e. succinate dehydrogenase (mitochondria), NADPH cytochrome c reductase (microsomes), 5Јnucleotidase (plasma membrane), and lactate dehydrogenase (cytosol)) and NMT activity (Table I and Fig. 1). Most (i.e. 60 -85%) of the original NMT activity was found in the subcellular fraction exhibiting the highest A 260 /A 280 and A 235 /A 280 absorbance ratios (i.e. Ϸ1.8 and Ϸ1.2, respectively), indicative of highly enriched ribonucleoprotein structures (45,46). Also found in the ribosomal fraction were relatively low levels of the mitochondrial, microsomal, plasma membrane, and cytosolic marker enzyme activities confirming the absence of gross contamination by other major subcellular fractions. These results demonstrate a colocalization of the majority of cellular NMT activity with the subcellular fraction most enriched in ribosomes. A preferential cofractionation of NMT activity with the ribosomal fraction was also found when similar experiments were carried out using MOLT-4 and HeLa cells (Fig. 1). When CEM cells were extracted and fractionated in homogenization buffer B containing 150 mM KCl instead of sucrose, the NMT activity in the ribosomal fraction was reduced from Ϸ78% to 52%, whereas the activity released into the cytosol increased from Ϸ16% to 23% (data not shown). When CEM ribosomal pellets (isolated in homogenization buffer A) were resuspended in buffer A containing either 150 or 300 mM KCl and recentrifuged at Ն100,000 ϫ g, Ϸ27% or Ͼ75%, respectively, of the NMT activity was recovered in the supernatant fraction with the balance of the activity remaining with the ribosomal pellet, indicating that relatively weak ionic interactions are involved in the association of hNMT with the ribosomes.
To exclude the possibility that NMT activity in the ribosomal fraction resulted from the association of the enzyme with some unidentified cosedimenting cellular fragment/organelle and/or macromolecular complex, the CEM ribosomal fraction was examined further by sucrose gradient centrifugation. In this analysis, the postmicrosomal supernatant fraction from CEM cells was subjected to centrifugation on a 12-25% isokinetic sucrose gradient, and the fractions collected from the gradient were assayed for NMT activity and analyzed for 260 and 280 nm absorbance (Fig. 2). As shown, most of the NMT activity was found to cosediment exactly with two prominent 260 nm absorption peaks consistent with an apparent association of NMT with ribosomal particle(s)/subunit(s). The identification of these absorption peaks as highly enriched ribosomal particle(s)/subunit(s) is supported by finding A 260 /A 280 and A 235 /A 280 absorbance ratios of Ͼ1.8 and Ն 1.2, respectively (data not shown) (45,46). These experiments provide the first demonstration of an association of hNMT with cellular ribosomes. Furthermore, these data, plus our subsequent identification of possible structural signals involved in targeting of hNMT to ribosomes (see below), provide new insights into the mechanism(s) regulating cotranslational N-myristoylation in mammalian cells.
Identification of a Ϸ60-kDa hNMT in the Ribosomal Frac-tions from CEM, MOLT-4, and HeLa Cells-The subcellular distribution of hNMT in CEM cells was also assessed by immunoblotting with an affinity-purified pAb raised against a a ND, not detected.

FIG. 1. Subcellular distribution of hNMT activity in human lymphoblastic leukemia and cervical carcinoma cells.
Human CEM, MOLT-4, or HeLa cells were lysed in homogenization buffer A containing 0.25 M sucrose and separated into nuclear (480 ϫ g for 20 min), mitochondrial (4,000 ϫ g for 20 min), microsomal (20,000 ϫ g for 30 min), ribosomal (Ն100,000 ϫ g for 90 min), and cytosolic (Ն100,000 ϫ g supernatant) subcellular fractions by differential centrifugation. The distribution of hNMT activity among subcellular fractions was determined in vitro by assaying for N-myristoylation of a synthetic N-glycylpeptide acceptor and presented as a percentage of total NMT activity.

FIG. 2. Cosedimentation of hNMT activity with CEM ribosomal particle(s)/subunit(s) during isokinetic sucrose density centrifugation.
The postmicrosomal supernatant fraction (20,000 ϫ g/30 min) from CEM cells homogenized in buffer A as described in Fig. 1 was applied to the top of a preformed 15-25% sucrose gradient and centrifuged at 50,000 rpm in a swinging bucket rotor for 90 min. Fractions (0.25 ml) displaced from the top of the gradient were collected and assayed for NMT activity, measured for absorption at 260 and 280 nm, and the percentage of sucrose was determined with a refractometer.
hNMT peptide (24). This analysis permitted the visualization of a number of lightly stained bands of various sizes in several of the subcellular fractions as well as one prominently stained band of Ϸ60 kDa which was seen only in the ribosomal fraction (Fig. 3, lanes 1-5). Preincubation of the primary antibody with 20 M peptide antigen before blotting completely prevented immunostaining of the Ϸ60 kDa band without diminishing the other minor bands, thereby establishing the identity of the Ϸ60 kDa band as the hNMT polypeptide (Fig. 3, lanes 7-11). Identical results were obtained by a similar analysis of hNMT in subcellular fractions from MOLT-4 and HeLa cells (data not shown). Immunoblotting of the ribosomal fractions from CEM, MOLT-4, and HeLa cells revealed hNMTs with nearly identical relative electrophoretic mobilities on SDS-PAGE, indicative of similar sized enzymes in all three human cell lines (Fig. 4, lanes [3][4][5]. A similar sized band was also immunostained in partially purified preparations of bNMT (24) (Fig. 4, lane 2), thereby establishing a close size identity of NMTs in human cells and bovine brain.
We reported previously that partially purified preparations of bNMT contain variable amounts of low molecular mass immuno-cross-reactive polypeptides (e.g. 46 -57 kDa) which appeared to arise primarily as a result of postextraction proteolysis of the NH 2 terminus of the Ϸ60-kDa bNMT (24). This proteolysis was shown to have no affect on the bNMT in vitro catalytic activity (24). To determine the susceptibility of hNMT to a similar NH 2 -terminal proteolysis, CEM ribosomal fractions were prepared in the presence or absence of proteolytic enzyme inhibitors, and the stability of the hNMT polypeptide was followed during storage at 4°C by immunoblotting. In the absence of proteolytic enzyme inhibitors, most of the Ϸ60-kDa hNMT was converted into a band of Ϸ46 kDa by 6 weeks (Fig.  5, lanes 2 and 3), whereas in the presence of inhibitors the native Ϸ60-kDa hNMT was unchanged even after 3 months (Fig. 5, lanes 4 and 5). These experiments reveal a sensitivity of hNMT to proteolysis which is directly analogous to the proteolytic cleavage of catalytically dispensable NH 2 -terminal residues from yNMT (49 -51), bNMT (24), and Drosophila NMT (37). These findings are therefore consistent with the sugges-tion that the NH 2 -terminal domain of NMT may be important for regulating N-myristoylation in vivo (24,50,52) and that NH 2 -terminal proteolysis may constitute a physiological process to uncouple the enzyme from such regulatory constraints (24,37).
Identification of the ORF Most Likely Accounting for the Ϸ60-kDa hNMT-Our finding of a Ϸ60-kDa polypeptide for the native hNMT was surprising considering that the ORF thought to account for hNMT predicts a substantially smaller (i.e. Ϸ48 kDa) enzyme (36). Although additional in-frame methionines were reported Ϫ183 and Ϫ186 nucleotides upstream (i.e. predicting a Ϸ55-kDa polypeptide) from the originally chosen start site, the smaller ORF was chosen as the most likely hNMT gene based upon homology with the yNMT gene and the observation that a recombinant Ϸ48-kDa hNMT corresponding to the chosen ORF was catalytically active. However, considering our finding of a substantially larger Ϸ60-kDa native enzyme in human cell lines (see Fig. 4), the existence of nonhomologous (both in length and sequence) NH 2 -terminal domains of yeast, fungi, and human NMTs (36,37,(52)(53)(54) and the fact that the Ϫ183 codon also conforms closely to predictions of a prospective start site (55)(56)(57), we speculated that the native hNMT could actually be translated from the larger ORF defined by one of the two adjacent upstream start site(s) (i.e. at codon Ϫ186 or Ϫ183). However, considering that this larger ORF predicts a polypeptide of only Ϸ55 kDa, compared with our finding of a Ϸ60-kDa native enzyme on SDS-PAGE and the absence of in-frame stop codons in the known sequence even further upstream, it was also possible that translation of hNMT could have actually initiated from start sites upstream from the published sequence. To identify possible in-frame start site(s) and/or stop codons upstream of the known 5Ј-sequence, we performed 5Ј-RACE polymerase chain reaction analysis on a human brain cDNA library using internal primers based on the known hNMT gene (36). Polymerase chain reaction generated one major cDNA consistent with the existence of a single mRNA for hNMT in human brain (data not shown). Subsequent sequencing identified two new in-frame stop codons Ϫ255 and Ϫ270 nucleotides up-stream from the originally assigned start site with no additional in-frame ATGs (Table II). The presence of these in-frame stop codons plus the absence of additional conventional start sites upstream of the known sequence establish that one of only two of the previously identified ORFs must account for the native Ϸ60-kDa hNMT.
We and others (24,35) have speculated that bNMT may be translated from an ORF analogous to the larger of two possible ORFs in the hNMT gene (36). This suggestion was based upon a close amino acid sequence homology between hNMT and bNMT (35) and the finding that although bacterially expressed hNMTs corresponding to the originally proposed ORF (36) exhibit apparent molecular masses close to the predicted size (i.e. Ϸ48 kDa) (58, 59), native bovine and human NMTs were found to migrate as Ϸ60-kDa polypeptides on SDS-PAGE (24,35). We have further examined the question of which ORF accounts for native hNMT by preparing bacterial plasmids containing inserts encoding amino-terminal FLAG epitope-tagged hNMTs corresponding to each of the two possible ORFs of the hNMT gene. As shown in Fig. 4 (lane 6), blotting the extracts from bacteria expressing a recombinant hNMT corresponding to the larger of the two ORFs with an anti-hNMT pAb immunostained a prominent polypeptide that migrates in a way similar to the native Ϸ60-kDa hNMT. In contrast, a recombinant hNMT corresponding to the smaller of the two ORFs migrates with an apparent molecular mass of Ϸ50 kDa (as shown below). To exclude the possibility that the product of the smaller ORF might migrate anomalously as a larger (i.e. Ϸ60 kDa) band when expressed in mammalian cells because of post-translational modification(s), eukaryotic plasmids expressing the same two FLAG epitope-tagged hNMTs were transfected into HeLa cells, and the respective translation products were identified in subcellular fractions by anti-FLAG immunoblotting (see Fig. 6). This experiment revealed Ϸ60 or Ϸ50 kDa bands (i.e. corresponding to the larger and smaller hNMT ORFs) in HeLa cells which were identical in size with the respective recombinant polypeptides expressed in bacteria. These results therefore excluded the possibility that the native Ϸ60-kDa hNMT results from the anomalous migration of the smaller gene product. Furthermore, our finding that the recombinant hNMT encoded by the larger ORF migrates in a way similar to the Ϸ60-kDa native enzyme supports the proposal that the hNMT gene is actually initiated from one of the adjacent start codon(s) upstream from the originally assigned start site, defining an ORF consisting of 1431/1434 nucleotides and encoding a 477/478-residue polypeptide with an apparent molecular mass of Ϸ60 kDa.
Identification of a Prospective Ribosomal Binding Signal in the Amino-terminal Domain of hNMT-Our previous finding that NH 2 -terminal proteolysis converted the native Ϸ60-kDa bNMT subunit into 46 -48-kDa forms with no loss of in vitro catalytic activity (24) was consistent with earlier suggestions that the NH 2 terminus of NMT may have noncatalytic regulatory function(s) (50,52). We have now used plasmids expressing in human cells recombinant FLAG-NMTs corresponding to the two possible ORFs of the hNMT gene (see Table II) to test the hypothesis that targeting of hNMT to the ribosomal subcellular fraction is determined primarily by sequences encoded within the NH 2 terminus of the Ϸ60-kDa enzyme. Expression vectors encoding the two FLAG epitope-tagged hNMTs were transfected into human HeLa cells, and after 24 h the postnuclear cell extracts were separated into mitochondrial, microsomal, ribosomal, and cytosolic fractions by differential centrifugation as described above and analyzed for expression of FLAG-NMTs by immunoblotting with an anti-FLAG mAb. Analysis of cells transfected with the vector containing an insert corresponding to the larger ORF revealed a prominent Ϸ60-kDa FLAG-NMT almost exclusively in the ribosomal fraction (Fig. 6, lane 4), whereas cells transfected with the vector containing the insert corresponding to the smaller ORF expressed a Ϸ50-kDa FLAG-NMT predominantly in the cytosol (Fig. 6, lane 10). These results indicate that sequences within the NH 2 -terminal 10 -12-kDa domain of the Ϸ60-kDa FLAG-NMT are necessary for optimal targeting of hNMT to the ribosomal subcellular fraction. Nevertheless, as the level of FLAG-NMT protein expression increased during longer posttransfection cell incubations, higher amounts of Ϸ50-kDa FLAG-NMT in the ribosomal fraction and Ϸ60-kDa FLAG-NMT in the cytosolic fraction were observed (data not shown). Low affinity or nonspecific interaction between the Ϸ50-kDa FLAG-NMT and the ribosomal fraction may explain the increased presence of the enzyme in the ribosomal fraction during protein overexpression (60). On the other hand, the apparent "spillover" of the Ϸ60-kDa FLAG-NMT into the cytosolic fraction when expressed at higher levels may reflect a limited number of specific binding sites for the enzyme in the ribosomal fraction. Apparent differences in the affinities of Ϸ60and Ϸ50-kDa hNMTs for the ribosomal fraction were also examined in vitro. Serial dilutions of affinity-purified and catalytically active 2 Ϸ60or Ϸ50-kDa FLAG-NMTs were incubated with CEM ribosomes, and the FLAG-NMT bound to ribosomal pellets was determined by anti-FLAG immunoblotting after centrifugation (Fig. 7A). Densitometry analysis indicated clear differences in the affinities of the two FLAG-NMTs for the ribosomal pellet (Fig. 7B). Furthermore, the Ϸ60-kDa FLAG-NMT binding curve appears to approach a saturable plateau, suggesting a limited number of enzyme binding sites on the ribosomes. Assuming that Ϸ50-kDa FLAG-NMT binding to the ribosomal pellet is mostly nonspecific, then subtraction of this nonspecific binding from the Ϸ60-kDa FLAG-NMT binding curve yields a difference plot for the larger enzyme, suggesting saturable binding with an affinity around 1-2 nM (Fig. 7B, inset). However, regression analysis of the difference binding curve indicated optimum convergence to sigmoidal functions suggestive of a more complex binding process. On the whole, these results are consistent with our above transfection experiments (see Fig. 6) and confirm a preferential association of the Ϸ60-kDa FLAG-NMT for the ribosomal fraction compared with the Ϸ50-kDa FLAG-NMT. DISCUSSION N-Myristoylation is a cotranslational event (14,19) presumably occurring on cytosolic free ribosomes rather than membrane-associated ribosomes since nascent polypeptides containing the N-myristoylation motif do not contain the signal or topogenic bearing sequences required for targeting the protein synthesis complex to the endoplasmic reticulum (16,61,62). Although previous studies generally support this supposition by confirming the presence of NMT in the cytoplasm, such analyses have failed to establish a direct association of the enzyme with ribosomes as hypothesized (19,63). We have now used differential centrifugation of isoosmotically disrupted cell extracts to establish an apparent specific association of hNMT with the ribosomal subcellular fractions from human lymphoblastic leukemia (e.g. CEM and MOLT-4) and cervical carcinoma (e.g. HeLa) cells. This conclusion is based first upon our finding that the bulk of the in vitro NMT activity and the Ϸ60-kDa hNMT polypeptide are associated with the subcellular fraction containing both the highest A 260 /A 280 and A 235 /A 280 ratios (e.g. Ͼ1.8 and Ն 1.2, respectively) as well as relatively reduced levels of other subcellular organelle marker enzymes (see Figs. 1 and 3 and Table I). These observations indicate a preferential concentration of the majority of the hNMT with the subcellular fraction enriched in ribosomes. Second, a direct physical association of hNMT with putative ribosomal subunit(s)/particle(s) is inferred from the exact cosedimentation of the NMT activity with the 260 nm absorption profile during isokinetic sucrose gradient centrifugation (see Fig. 2). Finally, the specificity of subcellular targeting is indicated by a demonstration that optimum binding of hNMT to ribosomes appears to require NH 2 -terminal amino acid residues encoded from start site(s) (i.e. codon Ϫ183 or Ϫ186) upstream from and in-frame with the start site originally proposed for the hNMT gene (36) (see Figs. 6 and 7 and Table II). At least one of these upstream codons (i.e. codon starting at Ϫ183) would be a favorable start site because of its 5Ј-position and its context (i.e. it contains both an A at position Ϫ3 and a G at position ϩ4) and thus is close to the optimal consensus sequence for initiation by eukaryotic ribosomes (55)(56)(57). The presumption that this additional 5Ј-sequence is in fact translated in vivo is supported by our finding an apparent molecular mass for the native hNMT corresponding closely to a Ϸ60-kDa recombinant polypeptide expressed from the upstream start site rather than to a Ϸ50-  Fig. 1) transfected for 24 h with plasmids expressing FLAG epitope-tagged hNMTs corresponding to either the longer (lanes [1][2][3][4][5] or the shorter (lanes 6 -10) of two ORFs associated with the hNMT gene (36) were analyzed by SDS-PAGE (10% Novex gel) and immunoblotted with an anti-FLAG mAb. Mitochondrial fraction (lanes 1 and 7), microsomal fraction ( lanes  3 and 8), ribosomal fraction (lanes 4 and 9), cytosolic fraction (lanes 5 and 10), and molecular mass markers (lanes 2 and 6) (see Fig. 3) are shown. kDa recombinant polypeptide expressed from the originally proposed gene (see Figs. 4, 6, and 7). Furthermore, our finding of identical relative electrophoretic mobilities for the recombinant Ϸ50-kDa polypeptide expressed in both bacteria and HeLa cells excludes the possibility that the native Ϸ60-kDa hNMT results from anomalous migration during SDS-PAGE of the smaller ORF product due to intrinsic polypeptide properties and/or as a result of post-translational modification(s). On the other hand, the expression in bacteria and HeLa cells of a Ϸ60-kDa recombinant polypeptide that is slightly larger than that predicted from the upstream ORF (i.e. Ϸ55 kDa) most likely results from the highly charged nature of the added NH 2 terminus (64, 65). Our experiments thus provide in vivo and in vitro subcellular targeting and recombinant expression data identifying a native hNMT gene product that is 10 -12 kDa larger than is generally assumed for the enzyme and thereby indicates that the native hNMT is translated from a start site upstream from the one previously assigned. Furthermore, the additional NH 2 -terminal residues encoded by this larger ORF apparently provide most, if not all, of the high affinity ribosome-targeting signal enabling hNMT to function cotranslationally during protein synthesis. It is also possible, however, that residues located within the COOH-terminal Ϸ50-kDa domain of the Ϸ60-kDa enzyme also contribute lower affinity signals to promote optimum binding of the native hNMT to ribosomes in vivo.
The marked differences in both length and sequence among the NH 2 -terminal domains of yeast, fungi, and human NMTs (66) plus the finding that the NH 2 -terminal domains of yNMT (51), hNMT 2 (50), bNMT (24), and Drosophila NMT (37) are not  2-7) or Ϸ50-kDa (lanes 10 -15) FLAG-NMTs and equal sized aliquots of the postmicrosomal fraction (20,000 ϫ g for 30 min) from CEM cells (see Fig. 1) were incubated for 10 min at room temperature and centrifuged at 230,000 ϫ g for 30 min at 4°C. The resulting ribosomal pellets were analyzed by SDS-PAGE and immunoblotted with an anti-FLAG mAb. Lanes 1,8,9, and 16, molecular mass markers (see Fig. 3). Panel B, the immunostained Ϸ60-kDa (q) and Ϸ50-kDa (E) recombinant FLAG-NMTs were quantified by densitometry imaging, and the resulting integrated areas were plotted as a function of the molar concentration of the respective recombinant FLAG-NMT in the incubation mixtures described in panel A. Inset, the Ϸ50-kDa FLAG-NMT ribosomal binding curve was subtracted from the Ϸ60-kDa FLAG-NMT binding curve to yield a difference plot ascribed to specific ribosomal binding (OE). Data from the difference plot were fit to a three-parameter Chapman model by nonlinear regression analysis (dotted line). required for in vitro catalytic activity are consistent with the proposal that these NH 2 -terminal sequences could provide species-specific regulatory signal(s) that nevertheless are necessary for N-myristoylation in vivo (24,50,52). Our finding of an apparent ribosomal targeting signal within the NH 2 -terminal 10 -12-kDa domain of hNMT indicates that such an in vivo regulatory signal does in fact exist. Furthermore, the sensitivity of hNMT ribosomal binding to 150 -300 mM salt suggests that interactions between the NH 2 -terminal residues of hNMT and cognate acceptor site(s) in the ribosomal fraction involve relatively weak electrostatic interactions, a characteristic consistent with both the highly charged nature of the hNMT NH 2 terminus and reminiscent of a similar salt sensitivity exhibited by other cotranslationally active enzymes (e.g. N-methionylaminopeptidase and N-acetyltransferase) (67). Of particular interest is the presence in the NH 2 terminus of the Ϸ60-kDa hNMT of a polylysine block encompassing amino acid residues 37-50 (i.e. KKKKKKQKKKKEKG) with a 64% identity to the most NH 2 -terminal of two basic stretches found in human and rat N-methionylaminopeptidases involving residues 34 -47 (i.e. AKKKRRKKKKS/GKG) (68,69). Three such basic stretches are also found in the NH 2 terminus of the eIF-2␤ subunit of the eukaryotic initiator factor (i.e. a protein that mediates binding of the initiator Met-tRNA to the 40 S ribosomal subunit and mRNA before the start of translation (70)) with the most NH 2terminal basic stretch involving residues 12-25 of human eIF-2␤ (i.e. MSKKKKKKKKPFML) (71) and residues 14 -25 of yeast eIF-2␤ (i.e. ALKKKKKTKKVIPD) (72) exhibiting a 50% identity with the hNMT basic stretch. In all cases the groups of NH 2 -terminal basic amino acids are interspersed with stretches rich in acidic residues which together are speculated to account for the protein-protein and protein-nucleic acid interactions facilitating the participation of these factors in protein synthesis (69,(71)(72)(73). Similar basic patches formed by distant lysine residues which become closely apposed on the hydrophilic face of an amphipathic ␣-helix have been shown to be critical to formation of a "double-stranded RNA binding motif" (74 -77) and conserved COOH-terminal lysines in the connecting loops and/or within ␣-helices of methionyl-and tyrosyl-tRNA synthetases are critical for binding the nucleotide CCAA-end sequence of tRNA (78). Thus, considerable precedence exists for the involvement of basic amino acid motifs in the interaction of nucleic acids, including tRNA and mRNA, with proteins involved in protein synthesis. It is thus possible that the formation of similar basic patches within the nonhomologous NH 2 -terminal domains of other NMTs by secondary folding and juxtapositioning of distant basic residues could provide analogous ribosomal targeting signals for NMT in these more primitive eukaryotic organisms.
Regardless of the nature of the subcellular targeting signal or the precise acceptor(s) responsible for localization of hNMT to the ribosomal fraction, we envision a cotranslational model that involves the positioning of NMT within the protein synthetic complex so as to be in close proximity to its prospective nascent polypeptide substrate and thereby ensuring a timely response when transiently prompted by an appropriate N-myristoylation consensus signal. We speculate that the efficiency of this process would be enhanced substantially by interactions of NMT with specific proteins (e.g. fatty acid acyl-CoA synthetase, acyl-CoA-binding proteins, etc.) which could provide ready access to a pool of the relatively rare cosubstrate, myristoyl-CoA, while protecting NMT from competitive interference by a presumably much larger pool of palmitoyl-CoA. Stable interactions between hNMT and other cooperative proteins as hypothesized in this model could explain the existence of high molecular weight forms of bNMT in bovine brain (23,24) and hNMT in cultured cell 2 extracts. Furthermore, an apparent requirement for the NH 2 -terminal domain of NMT for formation of high molecular mass bNMTs (24) and for the ribosomal targeting of hNMT suggests the involvement of this noncatalytic NH 2 terminus in subunit multimerization and/or interaction(s) with other cellular proteins (e.g. those involved in protein synthesis).
In conclusion, our data provide the first direct evidence for the existence of a specific regulatory function associated with the NH 2 termini of mammalian NMTs. We propose that this NH 2 -terminal signal plays an important role in the targeting of hNMT to the site of protein synthesis on free ribosomes, thereby facilitating the participation of the enzyme in the cotranslational N-myristoylation of proteins in mammalian cells.